SingSong: Google's AI Experts Create Music That Complements Vocal Performances
In Brief
The innovative technology known as SingSong harnesses deep learning techniques learning model to produce music that aligns even better with vocal renditions compared to prior systems.
The Google research team envisions that this technology can be instrumental in developing karaoke tracks not only for seasoned professionals but also for those just starting out, enabling them to discover musical backings that resonate with their unique vocal styles.
A game-changer in karaoke: AI researchers have uncovered a method to create music tailored for singers.
SingSong , a revolutionary approach for unconditional audio production to facilitate training on the separated (vocal, instrumental) pairs, thus allowing for conditional 'audio-to-audio' generation tasks. AudioLM Discover the Top 5 AI Music & Audio Generators for Creating Royalty-Free Tracks
which has been trained on an extensive collection of music. This enables the system to produce accompaniment that is perfectly in sync with the singer's vocal timing and rhythm.
The new system, by contrast, uses a deep learning model For the evaluation, participants listened to pairs of 10-second vocal-instrumental mashups featuring the same vocals (sourced from MUSDB18-test) while varying the instrumentals from different origins (ground truth,
, or baselines). The inquiry prompts listeners to indicate which pairing they feel offers a more musically fitting instrumental backdrop to the vocals. google models Explore the Top 7 AI Voice Generators and Voice Cloning Options for Text-to-Speech
| Recommended post: By leveraging intricate architectures of deep neural networks and |
SingSong’s Fresh Examples
, developers can generate harmonic accompaniments instantaneously for extended segments without latency. generative models The previous examples utilized professional voices from the MUSDB18 dataset. We’re also intrigued by the capacity of SingSong to empower anyone, regardless of skill level, to create music using their own voice. This exploration includes vocal samples from the Vocadito dataset, which features recordings of amateur singers made using everyday electronics.
Currently, the system is in its nascent phase of development. Although the researchers acknowledge that further enhancements are needed before it can be launched for commercial use, they are optimistic about its potential to transform the karaoke scene and assist amateur singers in finding suitable musical accompaniments.
The evolution of singing: How AI is paving the way for anyone to become a professional vocalist.
Read more related articles:
Disclaimer
In line with the Trust Project guidelines Damir leads a talented team as the product manager and editor at Metaverse Post, focusing on topics such as AI/ML, AGI, LLMs, the Metaverse, and Web3 technologies. His work garners the attention of over a million readers each month. With a decade of expertise in SEO and digital marketing, Damir’s insight has been featured in notable outlets like Mashable, Wired, Cointelegraph, The New Yorker, Inside.com, Entrepreneur, BeInCrypto, and others. As a digital nomad, he travels between the UAE, Turkey, Russia, and the CIS. With a bachelor's degree in physics, he attributes his success in the rapidly evolving digital landscape to the critical thinking skills honed during his education.
